The relationship of government and technology has been cast to the forefront in the past two weeks, with the official introduction of the Burr-Feinstein anti-encryption bill, comments made by a US Attorney about banning “import of open-source encryption software”, and two congressional hearings on technological issues: one by the committee on energy and commerce, and one by the committee on oversight and government reform. All of this points to a need for greater understanding of the issues surrounding strong encryption, both in the context of this debate as well as in the government at large.
Strong Encryption is Indispensable
Strong encryption is a technological necessity for building and operating computing and communication systems in the modern world. It is simply not feasible and in many cases not possible to design these systems securely without building in strong encryption at a fundamental level. We are seeing an increase in the attacks against computing and communication infrastructure, and there is no reason to believe this trend will stop in the foreseeable future. Simply put, strong encryption is indispensable.
To fully understand the issue, however, we need to explore the specifics in greater detail.
Role of Strong Encryption in Secure Systems
Strong encryption plays a vital role in protecting information in modern computing and communication systems. Cryptography deals with methods of secure communication over insecure channels. Because of the scale, the distribution, and the inherent physics of modern communication and computing technology, it is simply not feasible (and in many cases, not even possible) to design and deploy “secure” channels and computing devices.
For example, it would be prohibitively expensive to replace the telecommunications grid with physically secure and shielded land-lines; moreover, this physical security system would be so large as to require its own “secure” communication channels. Wireless communication, on the other hand, can’t be secured by physical means at all. Similarly, physically securing every computing device is not even remotely possible, particularly with the proliferation of mobile devices. Finally, strong encryption is critical for protecting systems from threats like malicious insiders, physical theft or assault, persistent threats, and attackers who are able to breach the outer defenses.
Even with physical security, there are still systems that inherently rely on strong encryption to function. Authentication systems, which provide a means of securely identifying oneself inherently depend on the ability to present unforgeable credentials and communicate and store those credentials in a manner that prevents theft. Basic authentication mechanisms rely on encryption to communicate passwords and store them securely. Advanced authentication mechanisms such as the Kerberos protocol, certificate authentication, and CHAP protocols incorporate strong encryption on a more fundamental level, relying on its properties as part of their design. These systems are especially high-value targets, as they serve as the “gatekeepers” to other parts of the system. If an attacker is able to forge or steal authentication materials, they can gain arbitrary access to the system.
Necessity of Increased Use of Strong Encryption
Despite several assertions in the ongoing debates of “rapidly advancing technologies” and “going dark”, strong encryption is nothing new. The methods and ciphers have existed for decades, and various protocols and technologies have been using them for the better part of twenty years. Indeed, in certain applications such as banking, medical, and payment processing, use of encryption is mandated by law. Even when there are no statutory requirements, strong encryption has been used for decades in many applications to mitigate the civil liability risk of data loss.
Prior to 2013, areas such as commodity operating systems, mobile devices, communication protocols, and cloud storage had been lagging behind the aforementioned higher-risk domains in terms of their use of strong encryption for security. This was driven largely by a lack of perceived need. However, the increasing interconnectedness of devices and systems coupled with a steady increase in the number, scope, and sophistication of cyberattacks, together with the increase in attacks sponsored by organized crime, corporate, and nation-state entities has driven vendors to build strong encryption into new products by default. This is not criminals “going dark”. Rather, it is the world-at-large reacting to an increasingly hostile climate by shoring up its defenses.
This strengthening of defenses is necessary; the data breaches of 2015 are quite literally too numerous to cite here and affected everything from major retailers to critical government systems. This trend is expected to continue if not increase. Because attackers tend to target the weak links in a system, we can expect systems that fail to employ strong encryption in their design to become targets for attacks. Moreover, because of the increasing interconnectivity of devices and sophistication of attacks, we can expect these systems to become entry-points for multi-stage attacks and persistent infiltration.
The Fallacy of Secure Back-Doors
The notion of a secure back-door or “golden key” is a theme that has surfaced again and again in the ongoing debate on encryption. Moreover, this notion played a central role in the similar debate that took place in the 1990’s.
In 1994, there was a push to legislate the Escrowed Encryption Standard (EES) as legally-usable crypto and to ban unescrowed encryption. The EES hardware implementation was named “Clipper”, and was designed to provide the very sort of back-door access to encrypted traffic that has been the subject of recent debates. This push lost its momentum when researchers discovered critical flaws in the cipher. A very recent attempt by the British GCHQ to design a similar cipher has been found to have similar flaws.
In the mid-2000’s, the NSA introduced a surreptitious back-door into the Dual-EC random-number generation standard. This back-door was designed to allow the NSA to reconstruct the stream of random numbers generated by the algorithm, thus allowing them to decrypt traffic. The vulnerability was speculated about and exploits were developed by third-party researchers, and it was ultimately revealed to be the result of a deliberate effort by the NSA in the Snowden documents. This back-door vulnerability has been a root cause of at least one high-profile breach: the Juniper ScreenOS vulnerability, which affected a number of high-security networks including the U.S. State and Treasury departments.
These real-world cases demonstrate the practical danger of back-doors. On a more abstract level, a “secure” back-door is a paradox for the simple fact that any back-door is inherently a vulnerability. Introduction of covert vulnerabilities into security systems has been one of the leading causes of exploits. Doing so introduces added complexity and anomalies that an experienced researcher can detect and ultimately find ways to exploit.
Moreover, even if a back-door could be engineered in such a way as to be undetectable, there still remains the problem of protecting the information necessary to exploit the back-door. Were back-doored encryption to be mandated by law, the information necessary to exploit it would be invaluable, as it would provide uncontrolled, unmitigated access into every system using the standard. We can and should expect rival nation-state entities to employ every means to steal this information and were they to succeed, the result would be a severe national security crisis.
There is a scientific consensus among security researchers that back-doors cannot be engineered in such a way that does not introduce severe security risks. Moreover, it is very telling that agencies such as GHCQ and the NSA have not produced such a system themselves, despite their considerable mathematical and computational resources and decided interest in doing so. To ignore these facts and attempt to mandate back-doors would introduce critical and systemic vulnerabilities and grave risks to U.S. national security.
The Futility of an Encryption Ban
Even if secure back-doored cryptography were possible and the access materials could somehow be kept secure from attackers, a ban on strong encryption would be futile for the simple fact that it could not effectively be enforced. It would be impossible to prevent anyone from obtaining the source code of, or at least the knowledge of how to implement strong crypto even within the U.S., let alone outside of it.
For starters, encryption software is ubiquitous. Strong crypto has been the subject of extensive academic research for over half a century and has been written about in dozens textbooks and thousands of research papers. Exact descriptions of strong encryption algorithms have been published in international standards by multiple bodies. There are many implementations of these algorithms in both open- and closed-source software used around the world. Moreover, these algorithms can be printed on a few sheets of paper, or even on a T-shirt.
Attempting to ban access to strong encryption is tantamount to attempting to ban the possession and implementation of widespread and pervasive knowledge. Banning knowledge is as futile as it is misguided, and even if it could work, it would apply only to U.S. persons. It would not prevent foreigners from obtaining and using knowledge about crypto. Moreover, there is a long history of case law that would render any such action unconstitutional. Griswold v. Connecticut arose from an attempt to ban possession or use of knowledge almost a century ago; more recently, Bernstein v. United States establishes the publication of open-source software as a form of free speech, protected by the First Amendment.
Lastly, even if such a ban could stand legally, strong encryption could still be utilized through the related technique of steganography which provides methods for surreptitiously embedding information inside seemingly innocuous data. As a simple example, a hidden, encrypted message or file can be disguised as ordinary background noise in an image. It is easy to see how this can be used to defeat any attempt to enforce a ban on encryption.
More fundamentally though, cryptography arises out of mathematics; it is not something we created, but rather something we discovered. Trying to control the laws of mathematics through legislation is a doomed effort. Rather, we should focus our efforts on finding ways to make the most of what encryption offers.
Impacts on the U.S. Infosec and Technology Industries
The U.S. information security and technology sectors rely on strong encryption to build secure products and maintain their competitive advantage. Any ban or restriction on the ability of U.S. companies to use strong encryption in their products will almost certainly have serious negative consequences for these sectors. This would likely lead to a serious negative impact on the U.S. economy and workforce, as well as national security and technological advantage.
Such a ban would amount to a guarantee that software produced inside the U.S. is insecure, which would create a critical competitive advantage for companies based outside the U.S. The inability to properly secure software would prevent the information security industry from being able to operate effectively, and we should expect to see those firms immediately begin relocating operations to foreign countries where no such ban exists. The competitive disadvantage imposed by being unable to produce secure software would likewise drive much of the software and technology sectors to move primary development activities off-shore, if at a slower rate. The end result of this would not be the sort of universal access by law enforcement that these policies seek to provide, but rather a world where secure software incorporating strong encryption is produced by foreign nations, but not within the U.S.
We can expect that this move by industry would be echoed in the workforce, with the best workers emigrating as soon as possible to avoid negative impacts on their careers, followed by larger migrations driven by a shrinking job pool. There is already a global shortage for technology workers, and several savvy nations have programs in place to encourage technology workers to immigrate, bringing their talents (and tax revenues) with them. We could expect more of these sorts of policies should U.S. policies turn against the infosec and technology sectors, as foreign nations seek to capture talent leaving U.S. This sort of foreign migration of an entire sector was evident during the 1990’s and early 2000’s, when export of strong crypto was controlled under arms trafficking laws within the U.S.
This risk to the information security and technology industry and the potential loss of the U.S.’s technological advantage was directly referenced during the energy and commerce hearing multiple times. The industry panel confirmed that this is a concern among the industry leaders. The law enforcement panel rebuffed the concern, but offered only a vague counterargument, stating that the demand for U.S. software would not be impacted because of the U.S.’s reputation. This argument, which asserts that general reputation will somehow override specific, serious, and material concerns about quality, is an example of magical thinking and does not reflect an accurate picture of how reputation works, particularly with regards to technology.
The impact of the loss of the U.S. information security and technology sectors, as well as the technological advantage enjoyed by the U.S. as a result of its position within these industries on the U.S. economy would be catastrophic. Moreover, the impact on national security would similarly be severe as it becomes necessary to look abroad for software vendors and security solutions. Policies that industry leaders agree are likely to lead to this scenario are simply not a risk the U.S. can afford to take.
Relationship of Government and Technology
On a broader scale, we are facing a problem rooted in the relationship of technology and government. The congressional hearings in particular point to a number of issues in this relationship, ranging from outdated systems, to lack of knowledge and understanding, to a generally disorganized approach.
Encryption is Complex and Requires New Thinking
One of the key difficulties of the issues surrounding encryption is the fact that it is very different from what existing laws and policies have grown accustomed to regulating. This became evident in the congressional hearings, with representatives and law enforcement officials proposing “real-world” analogies, which do not hold up under more serious scrutiny.
If we are to use analogies to think about encryption and information security, the only really appropriate one is the world of microbiology, where pathogens are ubiquitous, adaptive, and require constant suppression by various immunity mechanisms. An immune system in this environment is not an extra feature, but an absolute necessity for continued existence. In such an analogy, the most dangerous pathogens of all are those that target the immune system itself; thus, any additional vulnerability such as a back-door opens up the entire system to such an attack.
More specifically, computational and communication infrastructure is vulnerable to attack because it is automatic, fast, and removed from human judgment. Institutions like banks can institute security policies for access to assets like safe deposit boxes and vaults that rely on human judgment and that are not susceptible to mass exploitation. The same is not true of systems protected by encryption: human judgment is far too slow to be a part of any computing process, and attackers can often use exploits against large amounts of data before being detected.
Lastly, the civil rights implications of encryption cannot be overlooked. Encryption is quite rare among technologies in that it directly protects and supports basic freedoms in an environment that is far less friendly to those freedoms than the physical world in which we live. While private communications can be conducted and accurate attributions can be made in the physical world, neither of these things are possible over the internet without strong encryption. With a significant portion of public discourse having moved to computing-based platforms, technologies such as encryption play a key role in protecting basic freedoms. Moreover, strong encryption is vital for activists living in countries with oppressive governments, state censorship, and discrimination. We must be careful to ensure that advancing technology does not erode basic rights, and technologies such as strong encryption play a vital role in doing so.
Encryption is a complex subject that cannot be accurately represented by any “real world” phenomenon, and requires effort to understand enough to form effective policy. Moreover, it is subject to a “weakest link” principle that mandates considerable caution when developing both systems and the policies that govern them. However, it is essential that we take the time and effort necessary to develop this understanding.
Technological Deficiency of Law Enforcement: A Serious Problem
One of the overarching themes of the congressional hearings- particularly of the first one is the apparent technological incompetence of high-level law enforcement officials. This is a very serious problem, especially with attacks by state-sponsored hackers and organized crime on the rise.
The first panel in the hearing by energy and commerce was composed of high-level law-enforcement officials. As a whole, these officials demonstrated an apparent lack of knowledge of the basics regarding technology and information security. Their testimony was of a wholly different tone from some of the press we’ve seen in the course of this debate. We have seen technically dubious PR, such as the claims about “dormant cyber-pathogens” and the New York Times’ characterization of what sounds like a command-line interface as encryption software. This sort of malicious PR is no doubt designed to exploit false public perceptions formed from inaccurate depictions of hacking in movies and TV to make its point.
However, I do not believe that was what we saw in the energy and commerce hearing; rather, the law-enforcement officials seemed to be making a genuine testimony, but were simply lacking in the knowledge and competency necessary to make a coherent, factually-correct point. In one of the more serious examples, one of the panelists responded to a question about the role of encryption in protecting authentication with a comment that authentication was a “firewall issue, not an encryption issue”. This makes no sense technically (firewalls generally don’t manage authentication, while encryption is central in the design of authentication protocols), and points to a fundamental lack of understanding about how secure systems work. Another panelist suggested statutory limits on the complexity of passwords. Simply put, such a policy would be nothing short of an information security catastrophe.
This lack of competence shows in the solutions that were proposed by the panelists, which largely focused on attempting to break encryption outright, or else legislate weaknesses into security systems to facilitate this course of action. This kind of thinking is common among novices in information security; experienced, knowledgeable actors such as professional hackers do not work this way. A professional hacker would not attempt to break encryption, but rather would focus on circumventing it through measures such as capturing keys, capturing data in an unencrypted form, social engineering, persistent malware, and forensic analysis.
The appropriate response to this by technologists is not scorn and arrogance, but rather alarm and action. The testimony in this hearing is evidence of a critical vulnerability in our law-enforcement system and by extension an inability to deal with the very real threats posed by the security problem. This suggests that law enforcement is in desperate need of assistance to develop the necessary competencies to deal with these issues. The technology sector can and should make efforts to educate and inform law enforcement, and help develop alternatives that do not weaken our infrastructure and create serious economic and national security risks.
Lack of Consensus within the Government
More generally, the hearings demonstrate a critical lack of consensus within the government as to how to act. This division was evident among the panelists as well as the representatives questioning them. Some demonstrate good technical competence, and make technically sound recommendations; others quite plainly do not.
Unsurprisingly, the most technically-competent areas of the government take a position in favor of strong encryption. The NSA for example, has voiced support for strong encryption, as has the Secretary of Defense. Former NSA and DHS heads have likewise voiced support for strong crypto. A report cited during the oversight and reform panel recommends (among similar points) that the U.S. Government “should not in any way subvert, weaken, or make vulnerable generally-available commercial software.”
Large sections of the government remain dangerously behind both in terms of technical competence and the state of their systems. We of course have the technically unsound arguments in favor of the introduction of back-doors and other weaknesses in critical systems. The oversight and reform hearing also revealed that some areas of the government are running dangerously out-of-date legacy systems, even referencing COBOL and punched-card based systems. This is a serious problem in a world where state-sponsored hackers are on the rise.
To give credit where due, the Obama administration has begun to make moves to address this. The foundation of the U.S. Digital Service, which seeks to draw talent from industry to address the problems within the government is a step in the right direction. However, the congressional hearings suggest that we will need to step up these sorts of efforts significantly in order to address these problems effectively.
The Burr-Feinstein Anti-Encryption Bill
The Burr-Feinstein anti-encryption bill (formally, the “Compliance with Court Orders Act”) represents the wrong kind of thinking and policy on the issue of encryption. The bill mandates that any producer of encryption software must provide access to encrypted data on demand. While the bill does contain a strange provision stating that it does not mandate or prohibit any design feature, the fact remains that it is impossible to comply with its basic stipulations for any system which includes strong end-to-end encryption. In spite of its assertion, the bill does effectively prohibit the development and use of these technologies.
As previously discussed, should the bill pass, we should expect the consequences with regard the U.S. information security and technology industries, the U.S. economy and workforce, U.S. national security and technological advantage, and the ability to defend against increasing information security threats to be very bad. Moreover, the bill’s direction is very much out-of-sync with the recommendations and directions of the most technically competent parts of the government, and would likely undermine their ongoing efforts.
More generally, this bill is simply the wrong direction. This kind of legislation will not work, as it will not prevent the development of truly secure software outside the U.S., nor can it prevent the use of strong encryption by criminals, state-sponsored hackers, and other extralegal entities. It does nothing to address the critical lack of technological expertise by critical areas of the government, including law enforcement. It stands to seriously undermine ongoing and important efforts to strengthen our defenses against a rising tide of attacks, and moreover, it is not at all clear how to comply with the bill’s stipulations while maintaining compliance with existing information security requirements in areas like banking, healthcare, payment processing, and storage of classified data.
Conclusion: Towards Effective Policy
Even though the congressional hearings served to highlight a number of problems, the overall tone was one of Congress taking action- which I believe to be more or less effective action -to understand and address these issues. Moreover, it was apparent that some members of Congress do possess an astute grasp of the issues surrounding information security and encryption. Of course, the existence of measures such as the Burr-Feinstein bill and the other problems I’ve mentioned show that we have quite a way to go.
I believe there is a need for the technology sector to take a proactive role in helping to shape these policies. These issues are extremely complex, and we need to apply our expertise to the problems we are facing to find solutions that won’t cause serious damage to our economy and national security. There are a number of issues that need to be addressed, including the following:
- Make addressing the increasing number and sophistication of cyberattacks and vulnerabilities in our infrastructure a policy priority.
- Address the pervasive presence of vulnerabilities in software as a whole.
- Proactively replace vulnerable legacy systems and update outdated IT practices within the government.
- Education and training to address the technological deficiency apparent in law enforcement competencies.
- Develop techniques, guidance, and equipment to enable law enforcement to capture data in an unencrypted state.
- Better understanding of the fundamental constraints governing what is possible with regard to encryption and information security.
- Develop mitigation scenarios and techniques to deal with loss of critical infrastructure due to an exploit.
- Further encourage and facilitate interaction with industry experts to help the government address these issues effectively.
In closing, one of the most telling remarks in the congressional hearings was the statement by an industry panelist that the state of software security is “a national crisis”. A crisis of this kind calls for action, and it is critical that we take the necessary steps to understand the issues, so that we may address the crisis effectively.